princeton-vl/RAFT-Stereo

The problem encountered during training with my own dataset.

Closed this issue · 2 comments

May I ask how to fine-tune a pre-trained model using my own dataset, which includes three subfolders under a main folder, one for left views, one for right views, and another one for disparity maps? I'm not sure how to proceed with the training process, so I would appreciate any help. Thank you very much.

You'll need to create a new subclass of StereoDataset in stereo_datasets.py, similar to what is done for Middlebury, KITTI, etc. Also under def fetch_dataloader(args): you will need to add a case for your particular dataset.

Thanks